Lumen AI

Lumen AI helps Dutch municipal planning and housing teams investigate potential housing locations, compare development concepts, and understand what needs to happen next. The Netherlands’ housing shortage affects affordability, independence and access to opportunity. Finding space is only part of the challenge: understanding a location requires combining fragmented information about buildings, parcels, planning rules, permits and the surrounding neighbourhood. Starting in Deventer, Lumen brings that research into an interactive map. Users can ask questions in natural language, explore research candidates, inspect source-backed findings, and generate indicative housing concepts in 3D. They can then revise those concepts conversationally and compare calculated housing capacity, social housing allocation, parking and shared green space. Research and proposals can be exported for further discussion. Meta powers the investigation and conversational workflow. Typed tools retrieve evidence, while deterministic calculations validate geometry and calculate supported metrics. Sources and unresolved questions remain accessible. Lumen distinguishes research signals from confirmed vacancy, availability or permission to build. I built the AI extension as a solo developer during the hackathon, using my pre-existing open-source Lumen World map platform as a separately attributed dependency. The submission focuses on the new investigation, scenario and conversational workflows. This is an early research tool, not a replacement for planners or statutory assessments. My ambition is to keep developing it with municipal teams: making early housing investigations easier to understand, compare and move forward.